After nearly perfectly predicting the Cowboys' Week 1 victory over the Giants and their Week 2 loss in Kansas City, I was pretty far off in Week 3. I projected a final score of 24-20 in favor of Dallas, and they of course ended up blowing out St. Louis.
Predicting final scores is a tough business, so I have to be happy with predicting the winner in all three contests, despite the poor results last Sunday. Nonetheless, I think the model has shown some really good accuracy in the past.
If you recall, I'm using individual player projections to predict the final score of Cowboys games. The projections are developed using apps like this one at rotoViz that scan historic game data for "player comps." In short, the apps use an algorithm to find players similar to the subject we want to study playing against defenses comparable to the one he's facing.
In theory, the model should become more accurate as the season rolls along because we'll have more data. To start the year, the only numbers we could use were those from 2012. Now, we have a decent amount of data on both the Cowboys and Chargers, so it should be easier to find truly similar players. Right now, the model is weighting 2012 stats only half as much as 2013 stats.
With that in mind, let's take a look at Tony Romo's comps for this game. These are the quarterbacks who most closely resemble Romo's stats over the past year (with his 2013 numbers getting the most weight).
Before even aggregating those stats, we see some big-time numbers for Romo. Part of that is because he had a quality game last week, but a bigger part is the ineptitude of the Chargers' pass defense. Quarterbacks with similar numbers to Romo, which you can see includes Romo himself four times and Peyton Manning four times, have dominated defenses that have posted numbers similar to those from San Diego.
If his comps are any indication, take a look at Romo's chances of throwing for X number of yards.
These numbers are drastically different from what we saw in the first three weeks. Of Romo's 25 comps, only eight (32 percent) have thrown for under 250 yards. But what's really amazing is the ceiling production; an incredible 40 percent of Romo's comps topped 351 yards passing against Chargers-like defenses.
The model also gives Romo nearly a coin flip's chance of throwing for more than 300 yards. In comparison, it predicted just a 12 percent chance of crossing 300 yards against the Chiefs!
Final Player Projections for Cowboys
Using the same methodology, here are the final stat projections for the main skill players in this contest.
- QB Tony Romo: 25-for-40 for 304 yards (7.6 YPA), 2.24 touchdowns, 1.16 interceptions
In addition to the high yardage projection, Romo is also projected to throw 2.24 touchdowns. That's a really high number, actually, and a great sign for Dallas. Notice that the numbers are somewhat dependent on a lot of attempts, meaning opposing quarterbacks have thrown a lot on pass defenses as poor as San Diego's, which makes sense.
Also notice that Romo is projected to throw 1.16 interceptions, a high number. Again, that's probably the result of an expected jump in attempts.
- RB DeMarco Murray: 60 rushing yards, 3.7 receptions for 27 yards, 0.48 rushing touchdowns
These are modest numbers for Murray, which is somewhat surprising. He's again projected to catch a lot of passes, so the model is obviously weighing his 2013 stats rather heavily; Murray is currently on pace for 85 catches this season.
- WR Dez Bryant: 4.9 receptions for 73 yards, 0.46 touchdowns
Bryant's touchdown projection seems unusually low. The Chargers will surely double Bryant at times, but he should be able to take advantage of a banged up secondary.
- WR Miles Austin: 3.9 receptions for 63 yards, 0.25 touchdowns
If Austin can't play, rookie Terrance Williams would be an obvious downgrade. Williams' touchdown projection would likely be comparable to Austin's, however, since the first-year receiver is red zone-relevant.
- TE Jason Witten: 4.5 receptions for 50 yards, 0.29 touchdowns
Despite his two-touchdown performance in Week 1, Witten has historically been sub-par in the red zone for a tight end. He's converted exactly 30 percent of his red zone targets into touchdowns during his career, compared to 41.6 percent for Chargers tight end Antonio Gates.
Final Player Projections for Chargers
- QB Philip Rivers: 23-for-37 for 258 yards (6.97 YPA), 1.52 touchdowns, 1.08 interceptions
- RB Ryan Mathews: 48 rushing yards, 2.2 receptions for 17 yards, 0.24 rushing touchdowns
- RB Danny Woodhead: 21 rushing yards, 2.4 receptions for 21 yards, 0.12 rushing touchdowns
- WR Eddie Royal: 3.7 receptions for 44 yards, 0.38 touchdowns
- WR Vincent Brown: 3.0 receptions for 41 yards, 0.46 touchdowns
- TE Antonio Gates: 3.7 receptions for 39 yards, 0.42 touchdowns
Rivers' projection is great news for Dallas. Not only is he projected well below Romo in touchdowns, but his YPA is significantly lower as well. If we could use only one individual stat to predict the final outcome, the best would be YPA (or some version of it).
Projecting the Final Score [embedded_ad]
Using these stat projections, we can begin to project the final score. Adding Romo's passing touchdowns to Murray's rushing touchdowns, we get 2.72 offensive scores. With the possibility of a defensive/special teams score or a rushing touchdown from someone other than Murray, we can adjust that number to three touchdowns, or 21 points.
Over the past 16 games, kicker Dan Bailey has connected on 32 field goals – exactly two per game. Thus, our final projection for Dallas is pretty simple: three touchdowns plus two field goals equals 27 points.
With the touchdowns from Rivers and the Chargers' backs, plus the probability of other scores, we can safely project them at 2.18 touchdowns. Adding in kicker Nick Novak's per-game average of 1.44 field goals over the past 16 games, we get a final projection of 19.6 points for San Diego, which we can round to 20.
If the teams were to play this contest 10,000 times, the most common result would likely be somewhere around a 27-20 victory for Dallas. Playing in San Diego, the Chargers should be given a few points, leaving us with a final game prediction of …
Cowboys 27, Chargers 23